On Wed, Oct 26, 2022 at 05:18:02PM -0500, Eric Blake wrote:
> Allow this filter to emulate qemu's behavior of a hard disconnect on
> write attempts larger than 32M.
> ---
> .../nbdkit-blocksize-policy-filter.pod| 21
> tests/Makefile.am | 12 +-
>
On Wed, Oct 26, 2022 at 05:18:01PM -0500, Eric Blake wrote:
> We have already exposed nbdkit_shutdown() as a way for a plugin to
> request that nbdkit itself shuts down when convenient (taking away the
> current and all other client connections). But we lacked a way for a
> plugin (or filter) to
On Wed, Oct 26, 2022 at 05:18:00PM -0500, Eric Blake wrote:
> - if (conn->sockin >= 0)
> -closesocket (conn->sockin);
> - if (conn->sockout >= 0 && conn->sockin != conn->sockout)
> -closesocket (conn->sockout);
> + if (conn->sockout >= 0 && how == SHUT_WR) {
> +if (conn->sockin ==
On Wed, Oct 26, 2022 at 05:17:59PM -0500, Eric Blake wrote:
> Instead of using magic numbers, track connection status via an enum.
> While at it, convert functions whose return value is not used to
> instead be void. No semantic change.
> ---
> server/internal.h| 16 +--
>
It's confusing having patches 1 & 3 separate.
I can't help thinking this whole series would be easier to understand
if instead of building a list of Python commands as strings, we built
a list of "instructions", where an instruction is a qualified union
which (in OCaml) would be:
type
On Wed, Oct 26, 2022 at 04:15:59PM -0500, Eric Blake wrote:
> Slightly rearrange the code so that we can reuse the -c framework for
> executing the code snippets for -u, --opt-mode, and --base-allocation.
> Slightly changes the error message when a URI is invalid (which
> requires revamping